FlixRecs — Personalized Recommendation Engine

End-to-end movie/show recommendation system with collaborative filtering and neural ranking for streaming platforms.

What It Does

• SVD matrix factorization on MovieLens 1M for collaborative filtering baseline
• Two-tower neural recommender: separate user and item encoder towers
• Real-time ranking with learned relevance scores
• FastAPI serving: /recommend/{user_id}, /similar/{movie_id}
• Evaluation: RMSE, HitRate@K, NDCG@K, Diversity

Results (MovieLens 1M)

Model	RMSE	HitRate@10	NDCG@10
SVD	0.873	0.71	0.52
Two-Tower	0.831	0.78	0.59

Tech Stack

Python | PyTorch | scikit-learn | FastAPI | Pandas | NumPy | FAISS

Quick Start

git clone https://github.com/neuralasmi/FlixRecs
cd FlixRecs
pip install -r requirements.txt
python train.py --model two_tower --epochs 20
python -m uvicorn app.main:app --port 8000
curl http://localhost:8000/recommend/123